DeepSeek shakes up the market: Chinese AI model challenges US tech giants

 

 

 

The recent impact of DeepSeek, a low-cost artificial intelligence model developed in China, has shaken up the US market. President Donald Trump called the model a "wake-up call" for the country's industries. The furor caused by Artificial Intelligence (AI) has rekindled the debate on the efficiency and high costs of developing these technologies. 

“A wake-up call for our industries.” US President Donald Trump spoke out about the sudden success of DeepSeek, a low-cost Chinese model of artificial intelligence that has shaken the largest US technology companies.

The artificial intelligence assistant has a much less expensive model than those used by OpenAI with ChatGPT. Something that has generated doubts in the US market about the real value necessary for the operation of AI, since it calls into question the enormous amounts of money that US technology companies seek to spend on data centers and chips to drive advances in their models.

In this regard, Trump said that the progress of DeepSeek is good because it shows that it is not “necessary to spend so much money.”

For his part, OpenAI CEO Sam Altman called it an “impressive model.” Although he defended its artificial intelligence model. “Obviously, we will offer much better models and it is really exciting to have a new competitor,” he said.

Very quickly, on Monday, January 27, DeepSeek became the most downloaded free application in the AppStore, and thus, caused a stir in the American stock market.

In fact, that same day it caused a 17% drop in the shares of Nvidia, leader in the artificial intelligence chip market, which resulted in the loss of 593 billion dollars in its market value. A historic loss for any company in one day.

The situation seems to be leveling out on Tuesday, January 28, at least in European markets, with gains in utility and media stocks supporting the index.

Also on Tuesday, DeepSeek limited new user registrations after suffering “large-scale malicious attacks.” Although it later assured that it had identified the problem and that a solution was being implemented.

The thunderous entry of Chinese AI into the American scene raises big questions among experts in this technology and, at the same time, shakes the perception that the Chinese industry was years behind the American ones.

What is DeepSeek?

While DeepSeek caused a massive frenzy in the US only this week, the startup was founded in 2023 in Hangzhou, China, and launched its first large AI-based model later that year.

Its director is named Liang Wenfeng, and he had co-founded one of China's leading hedge funds focused on AI-powered quantitative trading.

A month ago, the company launched a new AI model that claimed to be on par with American models but at a lower price.

One reason it appears to be more affordably priced is that it is built with less expensive Nvidia chips. As AP reports, in 2022, the United States restricted sales of high-performance chips to China. By then, Liang's fund had amassed 10,000 of Nvidia's high-performance A100 graphics processor chips. However, the models were later built with H800 chips, which, although lower in performance, have managed to create a sophisticated AI model. 

Adding to the surge in interest in DeepSeek was a paper published last week that talked about another DeepSeek AI model called R1. It showed “advanced reasoning abilities” and was much cheaper than the model designed by OpenAI, o1.

“Deepseek R1 is AI’s Sputnik moment,” venture capitalist Marc Andreessen said in a Sunday post on the social platform X. His message referenced the 1957 satellite launch that sparked a Cold War space exploration race between the Soviet Union and the United States.

Why has it caused so much furor and fear in the US?

There are several reasons why DeepSeek's entry into the board has set off alarm bells for its American competitors.

First, it calls into question the multimillion-dollar investments made by the tech giants to develop their AI models. According to the EFE news agency, DeepSeek's developers assured that the model was trained for only 55 days with a budget of less than six million dollars. Something that is considerably less than that of the American giants.

Second, it highlighted how expensive the shares of some of these companies are in relation to the broader market.

Before Monday's liquidation, Nvidia shares were trading at almost 60 times the value of its earnings, compared with 22 for the entire S&P 500, according to LSEG data.

"The overweighting of these tech stocks in many investors' portfolios and the high concentration of these tech stocks in market indices was a significant and underappreciated risk issue," said David Bahnsen, chief investment officer at The Bahnsen Group.

This is happening in a context in which the AI ​​craze has driven a huge flow of capital into the shares of these companies and has driven the stock markets to historic highs.

And thirdly, it also calls into question the US strategy of restricting the sale of AI semiconductors designed in that country to China. On this, several experts have pointed out that the timing of the furor over DeepSeek is no coincidence: it came shortly after Trump's inauguration.

“Trying to show that export controls are useless or counterproductive is a really important goal of Chinese foreign policy right now,” said Gregory Allen, director of the Wadhwani Artificial Intelligence Center at the Center for Strategic and International Studies.

Last week, among the multiple orders signed by Trump was one in which the Administration pledged to “identify and eliminate” “loopholes in existing export controls.” It also came after the U.S. launched the Stargate initiative that seeks to invest nearly $500 billion over the next four years to build up to twenty new data centers to support AI projects in the country.